I've been trying to adaptively smooth some cluster images with csmooth,
and these can be fairly extended, typically 500*500 pixels or more. I've
been using the FFT convolution option in csmooth, with regions of 2^n by 2^n
pixels where n is an integer.
The problem is that to avoid edge artifacts I need to smooth a larger
region, say 1024*1024 pixels, but this seems to take excessive amounts of
time and sometimes runs out of memory on a Pentium 3 700MHz CPU, with
500M of RAM. Is this due to a bug in csmooth, and can it be avoided?
This archive was generated by hypermail 2b29 : Fri May 24 2013 - 01:00:04 EDT